4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
12 #include <linux/f2fs_fs.h>
13 #include <linux/buffer_head.h>
14 #include <linux/backing-dev.h>
15 #include <linux/writeback.h>
21 #include <trace/events/f2fs.h>
23 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
25 if (f2fs_inode_dirtied(inode, sync))
28 mark_inode_dirty_sync(inode);
31 void f2fs_set_inode_flags(struct inode *inode)
33 unsigned int flags = F2FS_I(inode)->i_flags;
34 unsigned int new_fl = 0;
36 if (flags & FS_SYNC_FL)
38 if (flags & FS_APPEND_FL)
40 if (flags & FS_IMMUTABLE_FL)
41 new_fl |= S_IMMUTABLE;
42 if (flags & FS_NOATIME_FL)
44 if (flags & FS_DIRSYNC_FL)
46 if (f2fs_encrypted_inode(inode))
47 new_fl |= S_ENCRYPTED;
48 inode_set_flags(inode, new_fl,
49 S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|
53 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
55 int extra_size = get_extra_isize(inode);
57 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
58 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
59 if (ri->i_addr[extra_size])
60 inode->i_rdev = old_decode_dev(
61 le32_to_cpu(ri->i_addr[extra_size]));
63 inode->i_rdev = new_decode_dev(
64 le32_to_cpu(ri->i_addr[extra_size + 1]));
68 static bool __written_first_block(struct f2fs_inode *ri)
70 block_t addr = le32_to_cpu(ri->i_addr[offset_in_addr(ri)]);
72 if (addr != NEW_ADDR && addr != NULL_ADDR)
77 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
79 int extra_size = get_extra_isize(inode);
81 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
82 if (old_valid_dev(inode->i_rdev)) {
83 ri->i_addr[extra_size] =
84 cpu_to_le32(old_encode_dev(inode->i_rdev));
85 ri->i_addr[extra_size + 1] = 0;
87 ri->i_addr[extra_size] = 0;
88 ri->i_addr[extra_size + 1] =
89 cpu_to_le32(new_encode_dev(inode->i_rdev));
90 ri->i_addr[extra_size + 2] = 0;
95 static void __recover_inline_status(struct inode *inode, struct page *ipage)
97 void *inline_data = inline_data_addr(inode, ipage);
98 __le32 *start = inline_data;
99 __le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
101 while (start < end) {
103 f2fs_wait_on_page_writeback(ipage, NODE, true);
105 set_inode_flag(inode, FI_DATA_EXIST);
106 set_raw_inline(inode, F2FS_INODE(ipage));
107 set_page_dirty(ipage);
114 static bool f2fs_enable_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
116 struct f2fs_inode *ri = &F2FS_NODE(page)->i;
117 int extra_isize = le32_to_cpu(ri->i_extra_isize);
119 if (!f2fs_sb_has_inode_chksum(sbi->sb))
122 if (!RAW_IS_INODE(F2FS_NODE(page)) || !(ri->i_inline & F2FS_EXTRA_ATTR))
125 if (!F2FS_FITS_IN_INODE(ri, extra_isize, i_inode_checksum))
131 static __u32 f2fs_inode_chksum(struct f2fs_sb_info *sbi, struct page *page)
133 struct f2fs_node *node = F2FS_NODE(page);
134 struct f2fs_inode *ri = &node->i;
135 __le32 ino = node->footer.ino;
136 __le32 gen = ri->i_generation;
137 __u32 chksum, chksum_seed;
139 unsigned int offset = offsetof(struct f2fs_inode, i_inode_checksum);
140 unsigned int cs_size = sizeof(dummy_cs);
142 chksum = f2fs_chksum(sbi, sbi->s_chksum_seed, (__u8 *)&ino,
144 chksum_seed = f2fs_chksum(sbi, chksum, (__u8 *)&gen, sizeof(gen));
146 chksum = f2fs_chksum(sbi, chksum_seed, (__u8 *)ri, offset);
147 chksum = f2fs_chksum(sbi, chksum, (__u8 *)&dummy_cs, cs_size);
149 chksum = f2fs_chksum(sbi, chksum, (__u8 *)ri + offset,
150 F2FS_BLKSIZE - offset);
154 bool f2fs_inode_chksum_verify(struct f2fs_sb_info *sbi, struct page *page)
156 struct f2fs_inode *ri;
157 __u32 provided, calculated;
159 if (!f2fs_enable_inode_chksum(sbi, page) ||
160 PageDirty(page) || PageWriteback(page))
163 ri = &F2FS_NODE(page)->i;
164 provided = le32_to_cpu(ri->i_inode_checksum);
165 calculated = f2fs_inode_chksum(sbi, page);
167 if (provided != calculated)
168 f2fs_msg(sbi->sb, KERN_WARNING,
169 "checksum invalid, ino = %x, %x vs. %x",
170 ino_of_node(page), provided, calculated);
172 return provided == calculated;
175 void f2fs_inode_chksum_set(struct f2fs_sb_info *sbi, struct page *page)
177 struct f2fs_inode *ri = &F2FS_NODE(page)->i;
179 if (!f2fs_enable_inode_chksum(sbi, page))
182 ri->i_inode_checksum = cpu_to_le32(f2fs_inode_chksum(sbi, page));
185 static int do_read_inode(struct inode *inode)
187 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
188 struct f2fs_inode_info *fi = F2FS_I(inode);
189 struct page *node_page;
190 struct f2fs_inode *ri;
193 /* Check if ino is within scope */
194 if (check_nid_range(sbi, inode->i_ino)) {
195 f2fs_msg(inode->i_sb, KERN_ERR, "bad inode number: %lu",
196 (unsigned long) inode->i_ino);
201 node_page = get_node_page(sbi, inode->i_ino);
202 if (IS_ERR(node_page))
203 return PTR_ERR(node_page);
205 ri = F2FS_INODE(node_page);
207 inode->i_mode = le16_to_cpu(ri->i_mode);
208 i_uid_write(inode, le32_to_cpu(ri->i_uid));
209 i_gid_write(inode, le32_to_cpu(ri->i_gid));
210 set_nlink(inode, le32_to_cpu(ri->i_links));
211 inode->i_size = le64_to_cpu(ri->i_size);
212 inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
214 inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
215 inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
216 inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
217 inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
218 inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
219 inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
220 inode->i_generation = le32_to_cpu(ri->i_generation);
222 fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
223 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
224 fi->i_flags = le32_to_cpu(ri->i_flags);
226 fi->i_advise = ri->i_advise;
227 fi->i_pino = le32_to_cpu(ri->i_pino);
228 fi->i_dir_level = ri->i_dir_level;
230 if (f2fs_init_extent_tree(inode, &ri->i_ext))
231 set_page_dirty(node_page);
233 get_inline_info(inode, ri);
235 fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
236 le16_to_cpu(ri->i_extra_isize) : 0;
238 /* check data exist */
239 if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
240 __recover_inline_status(inode, node_page);
242 /* get rdev by using inline_info */
243 __get_inode_rdev(inode, ri);
245 if (__written_first_block(ri))
246 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
248 if (!need_inode_block_update(sbi, inode->i_ino))
249 fi->last_disk_size = inode->i_size;
251 if (fi->i_flags & FS_PROJINHERIT_FL)
252 set_inode_flag(inode, FI_PROJ_INHERIT);
254 if (f2fs_has_extra_attr(inode) && f2fs_sb_has_project_quota(sbi->sb) &&
255 F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_projid))
256 i_projid = (projid_t)le32_to_cpu(ri->i_projid);
258 i_projid = F2FS_DEF_PROJID;
259 fi->i_projid = make_kprojid(&init_user_ns, i_projid);
261 f2fs_put_page(node_page, 1);
263 stat_inc_inline_xattr(inode);
264 stat_inc_inline_inode(inode);
265 stat_inc_inline_dir(inode);
270 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
272 struct f2fs_sb_info *sbi = F2FS_SB(sb);
276 inode = iget_locked(sb, ino);
278 return ERR_PTR(-ENOMEM);
280 if (!(inode->i_state & I_NEW)) {
281 trace_f2fs_iget(inode);
284 if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
287 ret = do_read_inode(inode);
291 if (ino == F2FS_NODE_INO(sbi)) {
292 inode->i_mapping->a_ops = &f2fs_node_aops;
293 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
294 } else if (ino == F2FS_META_INO(sbi)) {
295 inode->i_mapping->a_ops = &f2fs_meta_aops;
296 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
297 } else if (S_ISREG(inode->i_mode)) {
298 inode->i_op = &f2fs_file_inode_operations;
299 inode->i_fop = &f2fs_file_operations;
300 inode->i_mapping->a_ops = &f2fs_dblock_aops;
301 } else if (S_ISDIR(inode->i_mode)) {
302 inode->i_op = &f2fs_dir_inode_operations;
303 inode->i_fop = &f2fs_dir_operations;
304 inode->i_mapping->a_ops = &f2fs_dblock_aops;
305 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_HIGH_ZERO);
306 } else if (S_ISLNK(inode->i_mode)) {
307 if (f2fs_encrypted_inode(inode))
308 inode->i_op = &f2fs_encrypted_symlink_inode_operations;
310 inode->i_op = &f2fs_symlink_inode_operations;
311 inode_nohighmem(inode);
312 inode->i_mapping->a_ops = &f2fs_dblock_aops;
313 } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
314 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
315 inode->i_op = &f2fs_special_inode_operations;
316 init_special_inode(inode, inode->i_mode, inode->i_rdev);
321 f2fs_set_inode_flags(inode);
322 unlock_new_inode(inode);
323 trace_f2fs_iget(inode);
328 trace_f2fs_iget_exit(inode, ret);
332 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
336 inode = f2fs_iget(sb, ino);
338 if (PTR_ERR(inode) == -ENOMEM) {
339 congestion_wait(BLK_RW_ASYNC, HZ/50);
346 int update_inode(struct inode *inode, struct page *node_page)
348 struct f2fs_inode *ri;
349 struct extent_tree *et = F2FS_I(inode)->extent_tree;
351 f2fs_inode_synced(inode);
353 f2fs_wait_on_page_writeback(node_page, NODE, true);
355 ri = F2FS_INODE(node_page);
357 ri->i_mode = cpu_to_le16(inode->i_mode);
358 ri->i_advise = F2FS_I(inode)->i_advise;
359 ri->i_uid = cpu_to_le32(i_uid_read(inode));
360 ri->i_gid = cpu_to_le32(i_gid_read(inode));
361 ri->i_links = cpu_to_le32(inode->i_nlink);
362 ri->i_size = cpu_to_le64(i_size_read(inode));
363 ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
366 read_lock(&et->lock);
367 set_raw_extent(&et->largest, &ri->i_ext);
368 read_unlock(&et->lock);
370 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
372 set_raw_inline(inode, ri);
374 ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
375 ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
376 ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
377 ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
378 ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
379 ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
380 ri->i_current_depth = cpu_to_le32(F2FS_I(inode)->i_current_depth);
381 ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
382 ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
383 ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
384 ri->i_generation = cpu_to_le32(inode->i_generation);
385 ri->i_dir_level = F2FS_I(inode)->i_dir_level;
387 if (f2fs_has_extra_attr(inode)) {
388 ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
390 if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)->sb) &&
391 F2FS_FITS_IN_INODE(ri, F2FS_I(inode)->i_extra_isize,
395 i_projid = from_kprojid(&init_user_ns,
396 F2FS_I(inode)->i_projid);
397 ri->i_projid = cpu_to_le32(i_projid);
401 __set_inode_rdev(inode, ri);
402 set_cold_node(inode, node_page);
405 if (inode->i_nlink == 0)
406 clear_inline_node(node_page);
408 return set_page_dirty(node_page);
411 int update_inode_page(struct inode *inode)
413 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
414 struct page *node_page;
417 node_page = get_node_page(sbi, inode->i_ino);
418 if (IS_ERR(node_page)) {
419 int err = PTR_ERR(node_page);
420 if (err == -ENOMEM) {
423 } else if (err != -ENOENT) {
424 f2fs_stop_checkpoint(sbi, false);
428 ret = update_inode(inode, node_page);
429 f2fs_put_page(node_page, 1);
433 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
435 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
437 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
438 inode->i_ino == F2FS_META_INO(sbi))
441 if (!is_inode_flag_set(inode, FI_DIRTY_INODE))
445 * We need to balance fs here to prevent from producing dirty node pages
446 * during the urgent cleaning time when runing out of free sections.
448 update_inode_page(inode);
449 if (wbc && wbc->nr_to_write)
450 f2fs_balance_fs(sbi, true);
455 * Called at the last iput() if i_nlink is zero
457 void f2fs_evict_inode(struct inode *inode)
459 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
460 nid_t xnid = F2FS_I(inode)->i_xattr_nid;
463 /* some remained atomic pages should discarded */
464 if (f2fs_is_atomic_file(inode))
465 drop_inmem_pages(inode);
467 trace_f2fs_evict_inode(inode);
468 truncate_inode_pages_final(&inode->i_data);
470 if (inode->i_ino == F2FS_NODE_INO(sbi) ||
471 inode->i_ino == F2FS_META_INO(sbi))
474 f2fs_bug_on(sbi, get_dirty_pages(inode));
475 remove_dirty_inode(inode);
477 f2fs_destroy_extent_tree(inode);
479 if (inode->i_nlink || is_bad_inode(inode))
482 dquot_initialize(inode);
484 remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
485 remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
487 sb_start_intwrite(inode->i_sb);
488 set_inode_flag(inode, FI_NO_ALLOC);
489 i_size_write(inode, 0);
491 if (F2FS_HAS_BLOCKS(inode))
492 err = f2fs_truncate(inode);
494 #ifdef CONFIG_F2FS_FAULT_INJECTION
495 if (time_to_inject(sbi, FAULT_EVICT_INODE)) {
496 f2fs_show_injection_info(FAULT_EVICT_INODE);
502 err = remove_inode_page(inode);
508 /* give more chances, if ENOMEM case */
509 if (err == -ENOMEM) {
515 update_inode_page(inode);
516 dquot_free_inode(inode);
517 sb_end_intwrite(inode->i_sb);
521 stat_dec_inline_xattr(inode);
522 stat_dec_inline_dir(inode);
523 stat_dec_inline_inode(inode);
525 if (!is_set_ckpt_flags(sbi, CP_ERROR_FLAG))
526 f2fs_bug_on(sbi, is_inode_flag_set(inode, FI_DIRTY_INODE));
528 /* ino == 0, if f2fs_new_inode() was failed t*/
530 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
533 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
534 if (inode->i_nlink) {
535 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
536 add_ino_entry(sbi, inode->i_ino, APPEND_INO);
537 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
538 add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
540 if (is_inode_flag_set(inode, FI_FREE_NID)) {
541 alloc_nid_failed(sbi, inode->i_ino);
542 clear_inode_flag(inode, FI_FREE_NID);
544 f2fs_bug_on(sbi, err &&
545 !exist_written_data(sbi, inode->i_ino, ORPHAN_INO));
548 fscrypt_put_encryption_info(inode, NULL);
552 /* caller should call f2fs_lock_op() */
553 void handle_failed_inode(struct inode *inode)
555 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
559 * clear nlink of inode in order to release resource of inode
565 * we must call this to avoid inode being remained as dirty, resulting
566 * in a panic when flushing dirty inodes in gdirty_list.
568 update_inode_page(inode);
569 f2fs_inode_synced(inode);
571 /* don't make bad inode, since it becomes a regular file. */
572 unlock_new_inode(inode);
575 * Note: we should add inode to orphan list before f2fs_unlock_op()
576 * so we can prevent losing this orphan when encoutering checkpoint
577 * and following suddenly power-off.
579 get_node_info(sbi, inode->i_ino, &ni);
581 if (ni.blk_addr != NULL_ADDR) {
582 int err = acquire_orphan_inode(sbi);
584 set_sbi_flag(sbi, SBI_NEED_FSCK);
585 f2fs_msg(sbi->sb, KERN_WARNING,
586 "Too many orphan inodes, run fsck to fix.");
588 add_orphan_inode(inode);
590 alloc_nid_done(sbi, inode->i_ino);
592 set_inode_flag(inode, FI_FREE_NID);
597 /* iput will drop the inode object */